Traditional marketing often relies on static strategies that fail to capture dynamic customer behavior. This paper introduces an integrated framework to model and control the customer lifecycle, bridging the gap between empirical data and computational simulation. Using the Customer Personality Analysis dataset, we implemented a five-stage methodology. We first identified three distinct customer segments (At-Risk, Standard, High-Value) using Gaussian Mixture Models. To address the lack of longitudinal data, we calibrated a normative transition model based on customer inertia principles. Our analysis revealed that marketing effectiveness is highly state-dependent; notably, At-Risk customers exhibited a 33.5% lift when targeted with catalogs. Leveraging these insights, we optimized a State-Aware control policy that achieved a mean cumulative reward of 27.02, significantly outperforming uniform and recency-based baselines. Robustness checks confirmed this strategic advantage persists even under substantial perturbations of the transition matrix. The primary contribution is a validated "behavioral bundle'' of states, dynamics, and response functions, providing a rigorous foundation for future Agent-Based Modeling.
Ali Nasirzonouzi (Fri,) studied this question.